How do I cite marbits in my scientific work?

It is important to give visibility to the bioinformatics platform in papers, posters and presentations. We have a logo that you can include in your material, a website and an official name. Please cite it as: «High-Performance computing analyses were run at the Marine Bioinformatics Service (MARBITS) of the Institut de Ciències del Mar (ICM-CSIC) in Barcelona.»
If you received help from someone who is not in the authors list, please acknowledge as: “We thank Pablo S. / Vane B. / Ramiro L. for his/her help to run multiple/specific bioinformatics analyses”.

How can I access marbits from outside the local network?

You should first log in to the gatekeeper server triton. If you don’t have an account on triton, you or your supervisor should ask IT for one. The easiest way is to log in to tritonvia ssh and then ssh normally to marbits. Alternatively, or if you are planning on transferring files between your local machine and marbits you can build a tunnel to marbits.

How can I transfer files from my local machine to marbits?

If you are within the local network, you can use rsync or scp. The last one works much like cp. It is advisable that you do the transfer from your local computer. In that case do it like this:

    rsync -avz --progress <my_file> <marbits_username>@marbits:/home/<marbits_username>
    # or
    scp <my_file> <marbits_username>@marbits:/home/<marbits_username>

You can expand the path in your command to put your file at any location you have access to. In case of doubt you can consult man rsync or man scp.

If you are outside the local network please see how to «dig» a tunnel to marbits.

Alternatively, you can use a FTP program to transfer files between machines.

Your home at marbits is tiny and all your files are in /mnt/lustre/bio/users/your-user-name. You can make navigation accross the filesystem easier by creating soft links at your home (or wherever you want) as follows:

ln -s <target-of-the-link> <name-of-link>
# as in
ln -s /mnt/lustre/bio/users/psanchez/megaproject/ megaproject

This last example would create a link called megaproject in the directory where the command was issued, that would point to a directory called megaproject in the lustre filesystem. This link will behave just as if it was the actual directory.

I was a user in biocluster. Where are my data now in marbits?

If you had an account in biocluster and you had any data there before the cluster upgrade, you should be able to find all your stuff in the new lustre filesystem. Please read here.

What has changed from biocluster to marbits?

Besides the new hardware, we have moved from the Rocks cluster software (v6) to openHPC (v1.3.4). Jobs were submitted with qsub to the Sun (now Oracle) Grid Engine, SGE. Now they are submitted with sbatch to Slurm. We also moved from a NFS storage system to a Lustre filesystem. See more here.

How do I know my «bank» account for option --account

You probably have only one «bank» account. If that’s the case, it’s also your default account and you don’t really need to specify it in your sbatch scripts. Nevertheless,

sacctmgr show assoc user=your-user-name

will show you all the bank accounts associated to your marbits user.

My program dies just after it starts and I’m absolutely positive that it is installed, that my files are correctly formatted and my paths are neatly written. What is failing?

Are you trying to run it from the master node? Don’t do it! Your job will be mercilessly killed if it tries to get hold of too many resources. See how to run jobs at Marbits to find out how to do it.

I have several open sessions in marbits, but I have only one open terminal window. What are those?

If you run who or w and you see that you have many than one session that you are not currently logged on, it may have happened because you closed a terminal window and left, or your screensaver put your computer to sleep or something like that. You may close your session with crtl + D or exit or logout. The system will eventually kill those ‘zombie’ sessions, but you can manually do it by:

who  #look to the pts NUMBER of the session you want to close
pkill -9 -t pts/NUMBER